intermediateServing ~40 min

Build a minimal RAG pipeline

Stand up a vector store, ingest documents, and answer a grounded query.

Scenario

You want model answers grounded in your own documents.

Prerequisites
  • ·A running inference endpoint (see the vLLM lab)
  • ·A vector database (e.g. Qdrant)
Tasks
0 / 3 tasks
  1. bash
    kubectl create deploy qdrant --image=qdrant/qdrant:latest
kubectl expose deploy qdrant --port=6333

  3. Expected · The answer cites the retrieved source.

Success criteria
  • ✓Documents are indexed in the vector store
  • ✓A query returns a grounded, cited answer
Hints
  • ·Keep the vector store close to the serving model for low retrieval latency.
  • ·Evaluate groundedness before rollout.